Coupling BM3D with directional wavelet packets for image denoising

TL;DR

This paper introduces a novel image denoising method that combines directional wavelet packets with BM3D, iteratively enhancing denoising performance by leveraging their complementary strengths.

Contribution

It proposes a cross-boosting iterative algorithm that integrates directional wavelet packet-based denoising with BM3D, achieving superior results over existing methods.

Findings

Outperforms state-of-the-art algorithms in denoising quality

Effectively captures edges and textures in images

Achieves competitive or superior results in experiments

Abstract

The paper presents an image denoising algorithm by combining a method that is based on directional quasi-analytic wavelet packets (qWPs) with the popular BM3D algorithm. The qWPs and its corresponding transforms are designed in [1]. The denoising algorithm qWP (qWPdn) applies an adaptive localized soft thresholding to the transform coefficients using the Bivariate Shrinkage methodology. The combined method consists of several iterations of qWPdn and BM3D algorithms, where the output from one algorithm updates the input to the other (cross-boosting).The qWPdn and BM3D methods complement each other. The qWPdn capabilities to capture edges and fine texture patterns are coupled with utilizing the sparsity in real images and self-similarity of patches in the image that is inherent in the BM3D. The obtained results are quite competitive with the best state-of-the-art algorithms. We compare the performance of the combined methodology with the performances of cptTP-CTF6, DAS-2 algorithms, which use directional frames, and the BM3D algorithm. In the overwhelming majority of the experiments, the combined algorithm outperformed the above methods.